This document describes a student project to monitor water quality in fish farms using a spectrophotometer and IoT sensors. The project aims to measure water parameters like color, oxygen, pH and transmit data in real-time to detect issues. Students conducted experiments adding dyes to water samples and using the spectrophotometer to analyze absorption spectra and correlate intensities at specific wavelengths to color concentration. Multiple linear regression and machine learning models were used to predict color levels. The project plans to develop an autonomous smart buoy with embedded sensors to continuously monitor open water quality.

1. Internet of Things Course
Prof. Ioannis Chatzigiannakis, Andrea Vitaletti
Department of Computer, Control & Management Engineering, Università
di Roma La Sapienza
Anno Accademico 2020-2021
Dr. Roberto Bruzzese
Group Project Fish Farming Water Monitoring
Course of Internet of Things
Prof.: Ioannis Chatzigiannakis
Prof.: Andrea Vitaletti

2. Fish Farming Water Monitoring
Using Spectrophotometer for Underwater Quality
Analysis

3. Aquaculture 4.0
• Farmers traditionally performs periodic on site
measurements/inspections of water quality
• Traditionally Technical staff are composed by
veterinaries, biologists, chemists using hand held
instruments
• Aquaculture 4.0 means unmanned 24/24 hours
aquaculture monitoring
• Real time data and complete historical data base
• Triggering of alarms when hazardous situations and
immediate correction

4. Water Quality Parameters
• Fitoplankton and Zooplankton
• Oxigene percentage (optimum 80%-120%)
• Ph(6.5 – 9),Salinity (5g/L – 40g/L),Alkalinity(50-250mg/L)
• Waste : Droppings and uneaten feeds
• Oil
• Heavy metals (mercury), Dioxins, Polychlorinated
biphenils, Pesticides
• Nitrogen dioxide, Nitrogen Catabolites, Ammonia (0-0.03)
• Turbidity (suspended solids or plankton) less than 1 NTU not more than 5 NTU
• Temperature (optimum 25°C - 35°C)

5. Water Color Detection
So we have seen the water quality parameters. Now we are interested that the spectrophotometer
having a qualitatively and quantitatively unknown water sample is able to determine which quality
parameters are present and possibly in what percentage. This is why in our case we have
oversimplified. That is, we have moved the determination of the quality parameters to a single quality
parameter, taking the color of the water as an example. So if our spectrophotometer will be able to
determine the color of an unknown water sample and possibly the quantity of color or the color density
of that sample, we would have solved our problem. In other words, we could therefore solve the
problem of determining the presence in a sample of water of the presence of petrol for example and in
what quantity. Because the process will be the same.

6. Set up of the in-container Lab Experiment
Two types of experiments were performed. A first experiment by
adding to each tank containing 2 liters of water a gradually
increasing quantity of dye powder. We have tried first with blue,
then with red, then with a mixture of blue and red. Finally a
second experiment was done. By measuring the different light
radiationsfrom20different colors.

7. Result of First in-vitro Lab Experiment
fromthisfirstgraphobtained by adding increasingredvalues, we inferthatthe frequencies(410 nm and) 860 nm have a positivelinearcorrelationbetween the
measured intensityand the percentage of redinthe aqueous solution. Thereforeare the optimal frequenciesto detectthe concentrationof redcolor in the aqueous
solution

8. Result of First in-vitro Lab Experiment
We can determine some wavelength nanometer in the graphic that indicate that by increasing the color concentration
leads to an increase in the absorption of light at these wavelength, and this causes a decrease in the intensity of light
measured by the spectrophotometer detector

9. Result of Second in-vitro Lab Experiment

10. Normalization of measurements of the Lab set-up
• The intensity measured by the spectrophotometer
detector is dependent on the position of the detector
in the container and this can vary from one
measurement set-up to another. This is why it is
needed a normalization of signal to respect to pure
water.
• In order to normalize the value of the signal plotted on
the y axis it should be calculated by dividing the value
of spectra for pure water by the value of spectra for
coloured water.
• Y = log ((spectra for pure water)/ (spectra for coloured water))

11. Spectrum wavelenghts
• VIS : visible light sources 400-750 nm
• NIR : near infrared over 750 nm
• UV : ultra violet 340-400 nm
• MIR : mid-infrared over 2500 nm
400-430 Violet
430-500 Blue
500-570 Green
570-620 Yellow to Orange
620-670 Bright Red
670-750 Dark Red

12. Trasmittance Measured from water sample in closed space
https://www.laserfocusworld.com/lasers-sources/article/14168107/led-light-source-
enables-mobile-spectroscopy-for-industry-and-consumer-use
https://www.laserfocusworld.com/lasers-sources/article/14168107/led-light-source-
enables-mobile-spectroscopy-for-industry-and-consumer-use

13. Spectrum wavelengths optimal points
• It can be also noticed that , at some wavelengths, by
adding a dyes the intensity of light is higher , while at a
different wavelength the same add causes a decrease
of intensity. The decrease of intensity signal means
that at that wavelength the absorption is higher.
• Therefore the intensity data measured using these
wavelengths light sources can be used to achieve a
higher sensitivity to colour concentrations.

14. Experimental Design and Implementation
• In this experiment I have added a teaspoon of dyes
(blue color powder) from 1 to 5 , at each sensor
measurement trial
• Therefore we will focus only on optimal wavelengths
outcomes , and we will divide the classes in 1 to 5
• The class labels and measured intensity data are then
used as input for classification algorithms that result in
a classifier for identification of the levels of colours, say
low, medium , high

15. Colour Prediction Models using Multiple Linear Regression

16. Colour Prediction Models using Multiple Linear Regression

17. What is the Beer-Lambert Law ?
The Beer-Lambert law is a linear relationship between the absorbance and the concentration, molar absorption
coefficient and optical coefficient of a solution:

18. What is the Beer-Lambert Law ?
The Beer-Lambert law states that there is a linear relationship between the concentration and the absorbance of the
solution, which enables the concentration of a solution to be calculated by measuring its absorbance.
La mole è l'unità di misura della quantità di sostanza.

19. Another way to predict the colour concentration Beer-
Lambert Law for Red Colour ?
La mole è l'unità di misura della quantità di sostanza.

20. What is the Beer-Lambert Law ?
La mole è l'unità di misura della quantità di sostanza.

21. Spectrum Intensity distribution on water composite samples

22. Spectrum Intensity distribution on water composite samples

23. Spectrum Intensity distribution on water composite samples
Figure . Raman scatteringspectra ofwater
and ethanol solutionswith variousethanol
concentrationswithin the regionof CH and
OH stretchingbands.

24. Near Infrared spectroscopy
La mole è l'unità di misura della quantità di sostanza.
SSome of the biggest advances in the food industry are being made possible by technological innovations such
as ome of the biggest advances in the food industry are being made possible by technological innovations such
as near-infrared spectroscopy, wh, which can identify certain compounds like fat, sugar, water, and proteins in
food, leading to information about calorie content, freshness, and quality of food that can help consumers make
better choices. By analyzing the absorption spectrum of an unknown material
and matching this measurement with a database of known molecules, it is
possible to determine the presence and quantity of certain ingredients—for
example, the percentage of cocoa in a chocolate bar.ich can identify
certain compounds like fat, sugar, water, and proteins in food, leading to
information about calorie content, freshness, and quality of food that can help consumers make
better choices. By analyzing the absorption spectrum of an unknown material and matching this measurement with a
database of known molecules, it is possible to determine the presence and quantity of certain ingredients—for example, the percentage
of cocoa in a chocolate bar.

25. Near Infrared spectroscopy
Quantitative and qualitative analysis
Infrared (IR) light stimulates the molecules in a material, causing them to vibrate; due to
resonance, an atomic compound with a certain binding energy is stimulated by a photon
with the same energy. Depending on the type of atomic bond, certain fundamental
vibrations, harmonics, and combination vibrations can be measured in the IR range. A
characteristic atomic connection thus occurs in several wavebands, depending on whether it
is a fundamental or a harmonic vibration. In the near-infrared (near-IR), usually only
harmonics are found.
In terms of food analysis, one of the goals is to determine the percentage of corresponding
ingredients, such as sugar, fat, or water content. This method is called quantitative
spectroscopy. However, in most cases, the object to be analyzed does not consist of a
laboratory-pure mixture. As a result, the identification of the ingredients in complicated and
unknown substances can be difficult to pinpoint because the individual oscillations can
overlap in the overall spectrum. In contrast to the quantitative analysis mentioned above,
this is a qualitative analysis. The mathematical models used here are much more complex
and have to be geared to a reduced target quantity in order to achieve meaningful results.

26. On the path of mobile spectroscopy
As a result, there has not yet been a release of a handy spectrometer that
consumers can carry in their pockets to get an immediate, detailed analysis of
every piece of food they eat. Fortunately, however, the first steps toward this
reality are now being made. In the near future, a mobile handheld spectroscope
could ultimately be used to scan a wide range of materials—food, medicine,
even the human body—and analyze them in real time.
A spectrometer consists of four main building blocks: a light source, a detector,
optics, and a statistical mathematical model (so-called chemometrics
https://www.laserfocusworld.com/test-
measurement/spectroscopy/article/16562702/chemometrics-software-from-bw-
tek-adds-new-algorithm), which derives meaningful information from raw data.
The photoresponse of silicon-based detectors extends to just over 1000 nm; for
longer wavelengths, sensors based on gallium arsenide are needed. The 780–1000
nm spectral region crystallizes out as a meaningful measuring range because it
shows measurable characteristic curves for almost all key substances.
Near Infrared spectroscopy

27. Chemometrics
BWIQ is a multivariate analysis software package that analyzes spectral
data to find internal relationships between spectra and response data or
spectra and sample classes. It combines traditional chemometric methods
like Partial Least Squares Regression (PLSR) and Principal Component
Analysis (PCA) with a new proprietary adaptive iteratively reweighted
Penalized Least Squares (airPLS) algorithm.
Chemometrics in Spectroscopy, Book • Second
Edition • 2018
Chemometrics is the science of extracting information from chemical
systems by data-driven means. Chemometrics is inherently
interdisciplinary, using methods frequently employed in core data-
analytic disciplines such as multivariate statistics, applied
mathematics, and computer science, in order to address problems
in chemistry, biochemistry, medicine, biology and chemical
engineering.

28. The case of compound water samples
https://www.laserfocusworld.com/lasers-sources/article/14168107/led-light-source-
enables-mobile-spectroscopy-for-industry-and-consumer-use
https://www.laserfocusworld.com/lasers-sources/article/14168107/led-light-source-
enables-mobile-spectroscopy-for-industry-and-consumer-use
FIGURE 2. Absorption spectra of water and methanol under different
mixing ratios.

29. The Smart Buoy at Work
https://www.laserfocusworld.com/lasers-sources/article/14168107/led-light-source-
enables-mobile-spectroscopy-for-industry-and-consumer-use
This model may be subject to change

30. The Smart Buoy Sea Immersion
https://www.laserfocusworld.com/lasers-sources/article/14168107/led-light-source-
enables-mobile-spectroscopy-for-industry-and-consumer-use
This model may be subject to change

31. The Smart Buoy Components
https://www.laserfocusworld.com/lasers-sources/article/14168107/led-light-source-
enables-mobile-spectroscopy-for-industry-and-consumer-use
This model may be subject to change

32. Future Development
https://www.laserfocusworld.com/lasers-sources/article/14168107/led-light-source-
enables-mobile-spectroscopy-for-industry-and-consumer-use
This model may be subject to change
- Improve waterproofness of in vitro spectrophotometer
- Extend the model to other water quality parameters
- Refine machine learning module
- Improve anchorage system
- Improve connection and data transfer from Arduino to B-L072Z-
LRWAN1
- Connect B-L072Z-LRWAN1 with IC880A LORA Concentrator
- Other

33. Internet of Things Course
Prof. Ioannis Chatzigiannakis, Andrea Vitaletti
Department of Computer, Control & Management Engineering, Università
di Roma La Sapienza
Anno Accademico 2020-2021
Dr. Roberto Bruzzese
Group Project Fish Farming Water Monitoring
Course of Internet of Things
Prof.: Ioannis Chatzigiannakis
Prof.: Andrea Vitaletti
Grazie per la Vostra Attenzione!